Soil moisture forecast is of great importance for precision irrigation, which stretches water resources, thus encouraging conservation among farmers. On this premise, this study puts forth a model based on a Temporal Convolutional Network (TCN) for enhanced soil moisture forecasting. TCNs are selected for their natural approach of modeling long term temporal dependencies in the temporal data that is crucial to model change in soil moisture level due to changes in weather conditions and type of soil. The model then uses info from the soil moisture sensors, the weather forecast and remote sensing of the land surface. Data pre-processing steps such as, data cleansing, data interpolation, and data normalization makes the raw data fit for modeling. The TCN is then compared with RF and SVM, two other ML models achieving better accuracy, although with comparatively low MAE and RMSE values. This model is trained on the history data and evaluated through cross-validation in order of selecting the hyperparameters of the number of layer, and the rate of dilation of layers. Real-time prediction is done for automatic irrigation systems and afterwards irrigation decisions may be taken based on the soil moisture predicted by the model. The identified system reveals potential for increasing the utilization of water and crop yields and hence promoting sustainable and resource-optimizing farming systems. © 2025 Elsevier B.V., All rights reserved.